{"id":"eaa2ae28-dd7b-4398-b2f5-377e0581010e","arxiv_id":"2608.00314","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A weighted-mean-field attraction model shows that random noise stabilizes insect swarms by dissolving escaping pairs, with malaria mosquito swarms mapped to the stable edge.","lead":"This paper proposes that insect swarms stay together through attraction to a weighted average of neighbors, with sound falling off as distance to some power. The model predicts that random noise helps hold swarms together by stopping pairs from spiraling away, and places real mosquito swarms near the edge of falling apart.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Excluding short-range repulsion, based on <100 samples at 0–1 cm, could remove the pair-formation instability that noise is claimed to stabilize.","rationale":"I read the paper as making two claims: (1) the minimal model exhibits noise-enhanced stability via pair breakup; (2) fitted Anopheles parameters sit near the stability edge. The second claim is explicitly post-hoc and the reader already lowered the verdict to CONDITIONAL because of that. My independent concern targets claim (1) at its root. The model's interaction rule is the only place where biology enters; if the excluded short-range repulsion exists, the pair-formation instability that noise stabilizes may not exist in real swarms. The supplementary evidence against repulsion is underpowered (fewer than 100 samples at 0–1 cm), and the cited literature on mosquito collision avoidance suggests such a force is likely. I am not arguing the model is wrong—the numerical results are internally consistent and the analytic pair breakup calculation (Eq. 20) is a genuine derivation from stated assumptions. But a minimal model can be mathematically correct yet biologically irrelevant if a key interaction is missing. The concrete test is straightforward and would settle whether the phenomenon is an artifact of the missing repulsion. Therefore the correct verdict is the same as the reader's: CONDITIONAL—accept the model as a plausible minimal description, but require the repulsion check (or equivalent) before the general claim about insect swarms is taken as established.","tokens_in":17601,"tokens_out":9132,"duration_ms":88075,"concrete_test":"Re-run the phase diagram (Fig. 3) with a short-range repulsive term added to Eq. (1), e.g., an isotropic force F_rep = C (r_rep/r)^n r̂ for r < r_rep, with C and r_rep chosen to prevent separations below ~1–2 cm, using the same N, L, and noise protocol. If the noise-stabilized region (η≈0.1, γ≈2.4) disappears or shifts by more than the experimental error bars, the central claim is not robust.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central numerical claim—that noise stabilizes swarms by disrupting pair formation (abstract; 'Swarm stability analysis')—depends on the model omitting any short-range repulsive interaction. The equations of motion (Eqs. 1–4) contain only attraction to a weighted mean field with 1/r^γ weighting. For γ>2, nearest-neighbor attraction dominates and produces the pairs whose breakup is the proposed stabilization mechanism. If real mosquitoes have a short-range repulsive interaction, the pair state may be qualitatively different or absent, and the phase diagram in Fig. 3 could change materially. The paper's only evidence against repulsion is Supplementary Fig. 9, which shows the distribution of acceleration projected onto the nearest-neighbor direction is 'nearly symmetric around zero' at short ranges. The authors themselves note, 'Data become sparse in the 0–1 cm regime, with each distribution containing fewer than 100 samples.' With <100 samples, a modest repulsive bias could easily be missed. The cited reference [36] (Gupta et al.) shows mosquitoes integrate visual and acoustic cues for collision avoidance, which is functionally a short-range repulsion-like behavior. Because the no-repulsion assumption is load-bearing for the pair-formation mechanism, the claim that noise stabilizes the swarm is not yet robust to a plausible alternative interaction.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a two-dimensional agent-based model of insect swarming in which each individual moves at constant speed and is attracted to a weighted mean field of neighbors, with the weighting falling off as 1/r^γ, plus a uniform random heading noise of strength η. The authors report numerical phase diagrams showing stable swarms over a broad parameter region, and the counterintuitive result that for roughly 2<γ<4 the swarm is stable only in the presence of noise, because noise destabilizes the bound pairs of insects that would otherwise escape and disintegrate the swarm. An analytic calculation in the Appendix equates pair-formation and pair-annihilation rates to derive the lower stability boundary (Eq. 20), which agrees with the numerical simulations. The paper further maps six experimental swarms of Anopheles mosquitoes onto the model, obtaining γ=2.4±0.1 and η=0.08±0.01, and argues that these swarms sit near the predicted stability boundary, speculating on the biological advantages of edge-of-instability dynamics. The code is publicly available.","tokens_in":17889,"tokens_out":6756,"duration_ms":69362,"significance":"If the central claims hold, the paper would show that a very simple local interaction rule—attraction to a weighted mean field with no alignment and no repulsion—can reproduce three key features of insect swarms: cohesion, pair formation, and noise-enhanced stability. The analytic bifurcation calculation (Eq. 20) is a genuine mechanistic derivation rather than a phenomenological fit, and the paper makes a falsifiable prediction about the ineffectiveness of broadband acoustic noise for swarm disruption. The open-code policy and the attempt to connect the model to publicly available experimental data are additional strengths. However, the strength of the experimental claim is tempered by the model's restrictive assumptions and by the lack of uncertainty quantification on the phase boundaries.","major_comments":[{"comment":"The absence of short-range repulsion is load-bearing for the pair-formation instability that noise is claimed to stabilize. The only direct evidence against repulsion is Supp. Fig. 9, which the authors themselves state contains fewer than 100 samples in the 0–1 cm range; such sparse data cannot resolve a modest repulsive bias. Reference [36] also documents visual and acoustic collision-avoidance behaviors in mosquitoes. Please add a sensitivity analysis with a plausible short-range repulsive interaction (e.g., a soft-core repulsion with variable strength and range) and show that the 2<γ<4 noise-stabilized region and the analytic lower boundary (Eq. 20) remain qualitatively unchanged, or obtain additional near-field data. Without this, the core mechanism is not robust to a plausible alternative interaction.","section":"Supplementary Material I, Fig. 9; main text p.2"},{"comment":"The phase boundaries in Fig. 3 are drawn without confidence intervals or sensitivity analysis. The classification of a swarm as stable relies on fitting the exponent α in Eq. (5) over finite time, but no threshold or statistical uncertainty is given. The analytic lower boundary additionally uses hand-set scales a≈1.18R_s and r0≈2R_s (Appendix, p.8), and the sensitivity of η_c to these choices is not reported. This matters because the claim that the experimental swarms reside 'on the edge of a transition to instability' depends on the location of this boundary relative to the fitted (γ,η) values. Please provide a sensitivity analysis of Eq. (20) to a and r0, and indicate the uncertainty in the phase boundary.","section":"Swarm stability analysis and Fig. 3; Appendix, Eq. (20)"},{"comment":"The parameter mapping uses the intersection of two 10% agreement bands in the dimensionless correlation length and angular diffusion constant. When multiple intersection points exist, the mean is taken, but the spread and the number of intersections are not reported. The reported γ=2.4±0.1 and η=0.08±0.01 therefore do not include the systematic uncertainty of the mapping procedure, and the distance of the fitted parameters to the phase boundary is not quantified. Please report the full intersection regions and the resulting uncertainty in the edge distance; without this, the 'edge of instability' claim is not quantitatively supported.","section":"Comparison to experimental data; Supplementary Figs. 13–14"}],"minor_comments":[{"comment":"Please specify the time window over which α is fitted and the criteria used to classify α≈0, α≈1, and α≈2. In Fig. 2, add error bars or the number of independent realizations.","section":"Eq. (5)"},{"comment":"Typo: 'We my express' should be 'We may express'.","section":"Eq. (14)"},{"comment":"The variables x(t) and y(t) are introduced informally. A short definition of the geometry would improve readability.","section":"Pair stability near a swarm, Eq. (13)"},{"comment":"The term 'helical orbit' in the abstract and introduction conflicts with the two-dimensional model; the simulated pair orbits are planar circles. Consider clarifying that the helical appearance arises only when the vertical dimension is included.","section":"Conclusion and Discussion"},{"comment":"The velocity-alignment plot includes each individual's alignment with itself, which is identically 1. State whether the self-pair is excluded from the plotted average or discuss its effect on the short-distance bias.","section":"Supplementary Fig. 10"},{"comment":"The notation D_eff is used for angular diffusion, but the relation between D_eff and the model parameter η is only given later in Eq. (23). Please define all symbols at first use.","section":"Eq. (22)"}],"recommendation":"major_revision","confidential_remarks":"This is a thought-provoking paper with a transparent model and an analytic mechanism. The main concern is that the no-repulsion assumption, which is critical for the pair-formation route to instability, is supported only by very sparse data, and the phase boundaries are presented without uncertainty quantification. The authors should be encouraged to add sensitivity analyses and to temper the 'edge of instability' claim until the mapping and boundary uncertainties are characterized. The paper falls within the scope of the journal and is likely to interest its readership after these revisions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"I'll cut to it. The paper has one genuinely new result and one overreach. The new result is the phase diagram for a pure attraction-to-weighted-mean-field model: for 2<γ<4, intermediate noise stabilizes a swarm by breaking up the nearest-neighbor pairs that otherwise pull members out in helical orbits. The analytic lower bifurcation curve, Eq. 20, comes from equating pair formation and annihilation rates, not from fitting the simulation, and it tracks the numerical boundary. That is a solid, self-contained mechanism, and it gives a practical prediction: broadband noise is not a good swarm-disruption tool, narrow tones might be.\n\nThe paper also does well on the data side in spirit. The fit of γ≈2.4 and η≈0.08 across six Anopheles swarms is a nice demonstration of how one might infer interaction laws from trajectory statistics, and the value is plausibly acoustic. But the abstract says the swarms 'reside on the edge of a transition to instability,' and the text says 'demonstrate.' That is stronger than what was done. The two dimensionless measures used to fit γ and η are the same measures used to locate the swarms relative to the boundary. The edge is model-derived, so it is not strictly circular, but the claim would need a held-out dataset or at least error bars on the phase boundary to carry that weight.\n\nThe soft spots are real but addressable. The no-repulsion assumption is load-bearing, because the pair-formation instability is what noise stabilizes. Supplementary Fig. 9 is the only direct evidence, and the 0–1 cm bin has fewer than 100 samples. That is underpowered; a modest short-range repulsion could suppress or alter pair formation, and the cited collision-avoidance work [36] suggests some short-range behavioral repulsion exists. The stress test that flagged this is not fatal, but the robustness check should be done. The analytic boundary also relies on hand-set scales a≈1.18Rs and r0≈2Rs, and a π/2 breakup threshold; the visual agreement with numerics is reassuring but not quantified. And the reproducibility statement is weak: 'Available on GitHub' with no URL or commit hash is not acceptable in a paper that wants its numerics taken seriously.\n\nFor a reader, the value is the mechanism and the phase diagram, not the biological verdict. The paper belongs in the literature as a minimal model, and it deserves a real referee. I'd send it to review and ask for: (1) a repulsion sensitivity test, (2) error bars or ensemble statistics on the phase boundaries, (3) proper code release, and (4) a softer framing of the edge-of-instability claim. With those, it would be a nice PRL-type contribution.","headline":"A genuinely new noise-stabilization mechanism in a pure-attraction swarm model, with a plausible but overclaimed mapping to Anopheles swarms; worth refereeing after robustness checks.","tokens_in":18407,"tokens_out":3998,"would_cite":true,"duration_ms":40500,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Noise stabilizes insect swarms by breaking up pairs, a numerical model argues.","keywords":["insect swarms","collective motion","noise-induced stability","pair formation","weighted mean field attraction","acoustic communication","Anopheles coluzzii","phase transition"],"falsifier":"A decisive test would be to add broadband random acoustic noise to a laboratory Anopheles swarm and measure the pair-escape rate and swarm variance: if stable swarms require noise to break pairs, an intermediate noise level should reduce pair escapes relative to silence; if instead noise monotonically destabilizes the swarm or pair escapes increase, the central claim fails. A second, cheaper falsifier is to repeat the 0-1 cm nearest-neighbor acceleration analysis with far more than 100 samples; a clear negative bias (repulsion) at short range would invalidate the no-repulsion assumption that t","tokens_in":1280,"feed_emoji":"🦟","tokens_out":3391,"duration_ms":64929,"temperature":0.7,"pith_summary":"This paper tries to establish that a swarm of flying insects can be explained by one simple rule: each insect turns toward the weighted center of mass of its neighbors, with nearer neighbors counted more heavily. The authors argue that this rule alone—no repulsion, no velocity alignment—reproduces stable swarms, Gaussian density profiles, and the observed phenomenon of pairs of insects spiraling away from the swarm. Their central surprise is that random noise helps: in an intermediate regime (roughly 2<γ<4), adding jitter breaks up these escape pairs and holds the swarm together, so a noisy swarm is more stable than a deterministic one. They back this with a two-body calculation that balances pair formation against noise-driven pair breakup, producing a bifurcation curve that matches numerical phase diagrams. Fitted to six Anopheles mosquito swarms, the model returns γ=2.4 and low noise, placing the swarms on the stable side of a transition to instability—an edge position the authors speculate is biologically useful.","feed_headline":"Noise stabilizes insect swarms by breaking up pairs","feed_subtitle":"An attraction-only model fits real mosquito swarms and places them at the edge of collapse.","key_machinery":"The central object is the weighted mean-field vector (Eq. 1): each insect accelerates toward the density-weighted centroid of all other insects, with the exponent γ controlling how strongly near neighbors dominate. Constant-speed, planar equations of motion (Eqs. 3-4) reduce the dynamics to heading angles. The paper's analytic core is a two-body pair-formation rate (Eq. 16) and a noise-driven pair-annihilation rate (Eq. 19); equating them gives the critical noise strength (Eq. 20) that marks the lower stability boundary. Pair escape is governed by the critical angle (Eq. 13) where attraction to the partner equals attraction to the rest of the swarm.","core_discovery":"The paper claims that a swarm of insects can be held together by nothing but mutual attraction to a weighted mean field of neighbors, with signal strength falling off as 1/r^γ. At γ>2, nearest-neighbor attraction pulls insects into tight orbiting pairs that peel away from the swarm; added stochastic noise randomly kicks pair members apart, letting them rejoin the swarm. Balancing the rate at which pairs form against the rate at which noise annihilates them yields a bifurcation curve (Eq. 20) that separates stable from unstable swarming and matches numerical phase diagrams. Fitting the model to recordings of six Anopheles coluzzii swarms gives γ=2.4±0.1 and η=0.08±0.01, placing the swarms jus","pith_inferences":["A direct, testable prediction follows from the pair-breaking mechanism: adding controlled heading noise to a laboratory swarm should reduce the rate of pair-escape events, which can be checked against existing tracking data.","The balance-of-rates argument suggests a generic criterion for any aggregation model with local attraction: stability is set by the ratio of pair-escape probability to pair-mixing rate, so other taxa with similar acoustic or visual attraction may also show noise-enhanced cohesion.","The fitted edge-of-instability position hints that swarms may exploit criticality; an editor-level extension is that perturbation response (e.g., to a passing female or a tone) should be anomalously large and long-range if the swarm is truly at the edge, which is testable.","The 2D reduction assumes the vertical direction is passive; in real swarms, vertical escape could alter pair trajectories, so extending the model to 2.5D would test whether noise stabilization survives."],"forward_implications":["If correct, swarming requires no velocity alignment or short-range repulsion; attraction to a weighted mean field with inertia is enough.","Random noise can be a stabilizing force: in the intermediate regime 2<γ<4, adding heading jitter keeps swarms bound by disrupting pair formation.","Pair formation (helical orbiting pairs) is predicted to occur transiently whenever γ>2, matching observed midge behavior.","Mapping experimental data to the model yields γ≈2.4, consistent with acoustic near-field signals decaying as 1/r^3 with monopole/dipole contributions; real swarms sit near the stability boundary.","Broadband acoustic noise is predicted to be ineffective at disrupting mosquito swarms; tonal or narrowband signals are a better candidate."],"fun_headline_variants":["Noise keeps insect swarms together by splitting pairs","Attraction alone explains swarm cohesion, noise prevents breakup","How random kicks save insect swarms from fragmentation","Mosquito swarms cling to stability via noise-stabilized pairs","Noise halts pair escape, keeping insect swarms stable"],"cache_read_input_tokens":19712,"weakest_assumption_plain":"The model assumes insects respond only by turning toward a weighted mean field of all neighbors, with no short-range repulsion, no velocity alignment, and no delay; if real swarms have a meaningful short-range repulsion or vertical escape route, the phase diagram and the noise-stabilization mechanism could change.","fun_headline_variants_meta":{"raw":{"variants":["Noise keeps insect swarms together by splitting pairs","Attraction alone explains swarm cohesion, noise prevents breakup","How random kicks save insect swarms from fragmentation","Mosquito swarms cling to stability via noise-stabilized pairs","Noise halts pair escape, keeping insect swarms stable"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000153,"raw_usage":{"total_tokens":1015,"prompt_tokens":687,"completion_tokens":328,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":431,"completion_tokens_details":{"reasoning_tokens":262}},"tokens_in":431,"tokens_out":328,"duration_ms":3927,"temperature":1.0,"reasoning_tokens":262,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T00:44:34.123890+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive test would be to add broadband random acoustic noise to a laboratory Anopheles swarm and measure the pair-escape rate and swarm variance: if stable swarms require noise to break pairs, an intermediate noise level should reduce pair escapes relative to silence; if instead noise monotonically destabilizes the swarm or pair escapes increase, the central claim fails. A second, cheaper falsifier is to repeat the 0-1 cm nearest-neighbor acceleration analysis with far more than 100 samples; a clear negative bias (repulsion) at short range would invalidate the no-repulsion assumption that t","supporting_citations":[],"review_version":1}